• en

• vision-language
• multimodal
• uncensored
• gguf
• text-generation
• image-understanding

• Gemma-3-4B

Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking-GGUF

This repository contains Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking-GGUF, a 4B-parameter vision-language instruction-tuned model provided in GGUF format for efficient local inference.
The model is designed for open-ended reasoning, multimodal understanding, and minimal alignment constraints, making it suitable for experimentation, research, and advanced local deployments.

Model Summary

• Model ID: Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking-GGUF
• Architecture: Gemma 3 (4B parameters)
• Type: Vision-Language (Text + Image)
• Format: GGUF
• Publisher: mradermacher
• License: Apache 2.0 (inherits from base model)

Key Characteristics

• Multimodal input support (text + images)
• Instruction-tuned for conversational and reasoning tasks
• Reduced content filtering and alignment constraints
• Optimized for local inference runtimes
• Suitable for research, exploration, and advanced user workflows

⚠️ This model is uncensored. Outputs may include sensitive or unfiltered content. Use responsibly.

Supported Use Cases

Text-Based

• Conversational assistants
• Creative writing and storytelling
• Summarization and rewriting
• General reasoning and analysis

Vision + Text

• Image captioning
• Visual question answering
• Scene and object understanding
• Multimodal reasoning tasks

GGUF Compatibility

This model can be used with GGUF-compatible runtimes such as:

• llama.cpp
• Ollama (GGUF-based builds)
• Other local inference engines supporting GGUF

Performance and supported features may vary depending on runtime and hardware.

Basic Usage Example

Command Line (llama.cpp-style)

./main \
  -m Andycurrent/Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-ThinkingGGUFF16.gguf \
  -p "Describe the key idea behind multimodal AI models."

Usage Notes

• Provide clear, explicit prompts for best results
• When using images, ensure proper formatting and resolution
• Add moderation or filtering layers if deploying in public-facing applications

Ethical Considerations

Due to its uncensored nature:

• Not recommended for unrestricted public deployment
• Should not be used in safety-critical environments
• Users are responsible for compliance with applicable laws and policies

Acknowledgements

• Gemma base model contributors
• Open-source inference and quantization communities
• Tools and runtimes enabling efficient local LLM deployment